CS2 Research Tests Player Fingerprints to Link Smurf Accounts
An NTNU master's project has tested whether mouse and keyboard patterns in CS2 demos can link accounts controlled by the same player, offering a possible new signal against smurfing and ban evasion.
A player fingerprint built from CS2 demos
A master's project at the Norwegian University of Science and Technology has explored whether Counter-Strike 2 demos can identify the player behind an account. Researcher Christopher B. Didriksen built a system that compares patterns in mouse movement and keyboard input to create what he calls a player fingerprint.
The method uses information already recorded in match demos. It does not require a new game client, kernel-level driver or additional sensor data. Each demo provides a sequence of view movements and button presses, which the system converts into separate mouse and keyboard profiles.
How the matching system works
Didriksen argues that repeated actions such as flicks, counter-strafes, recoil control and utility use become habitual over time. The system looks for these recurring patterns across different matches and accounts.
A fast first stage searches for accounts with similar profiles. Candidate matches then face a stricter confirmation step in which the mouse and keyboard signals must agree independently. A strong result from one input cannot compensate for a weak result from the other.
According to the researcher's published summary, the mouse profile selected the correct player in every comparison within the test data, while the keyboard profile did so in 98 percent of cases. Combining both signals separated all verified same-player pairs from unrelated accounts in that dataset.
The test uncovered previously unreported smurfs
The project initially included eight known pairs of linked accounts. Several results that appeared to be false positives were investigated manually and turned out to be additional smurf accounts that participants had not reported.
After those links were added, the researcher said the system identified all 13 verified same-person pairs in the dataset. One group connected four accounts to the same player even though they had been submitted under separate details.
These results are promising, but they should not be treated as proof that the method is ready for automatic bans. The number of verified linked-account pairs remains small, and the system has not been tested across the full CS2 population. The thesis itself has also not been released publicly for independent review.
Why the idea matters for competitive integrity
Most anti-cheat penalties apply to an account. A banned player can often return through a new account, while hardware, network and contact-information checks can be bypassed or shared by legitimate users. A behavioral match could give matchmaking platforms another way to recognize repeat offenders.
Didriksen suggests using an account link as supporting evidence rather than an automatic conviction. A platform could place the new account on a watchlist, require an additional signal or combine the match with an existing anti-cheat finding before extending a penalty.
The same approach could also help detect smurfing, account sharing or account sales. However, shared accounts create a major limitation because a single account would contain patterns from more than one person. New accounts also need several matches before the system has enough data for a confident comparison.
Valve has not adopted the system
The researcher says the prototype has been optimized to compare new demos against a very large account database and is seeking access to a real-world population for further testing. Valve has not announced a partnership, trial or integration with VAC or Trust Factor.
Privacy and appeal procedures would also need careful design before any deployment. The project consulted Norwegian university data-protection specialists and limited retention of demos and fingerprints, but large-scale use would still require transparent rules, security controls and a way to challenge incorrect links.
For now, the work is an experimental proposal with encouraging small-scale results. Its next meaningful test would be independent evaluation on a much larger and more diverse set of players.
